Drag & Drop to Analyze

Upload one or multiple invoices, receipts, or financial documents

Supports PDF, DOCX, XLSX, CSV, PNG, JPG, GIF, BMP, TIFF, WebP up to 25MB

100% Client-Side — Your files never leave your browser
100% Private
Runs in Your Browser
No Data Uploaded
PDF, DOCX, XLSX & Images

Quick Overview

Summary: ArmorValid detects forged, tampered, and AI-generated financial documents entirely in your browser. Drop a PDF, DOCX, XLSX, CSV, or image and get a risk verdict in about 30 seconds. We run 7+ forensic checks per document — covering metadata, arithmetic, pixel-level anomalies, and statistical distribution — and never upload your files.

  1. Upload — drop a file (up to 25 MB) or a batch of up to 25 files.
  2. Analyze — OCR, Error Level Analysis, metadata forensics, arithmetic validation, Benford's Law, EXIF, and SHA-256 hashing all run locally.
  3. Review — get a verdict (Genuine, Suspicious, or High Risk), weighted risk factors, anomaly overlays, and a downloadable PDF or JSON report.

Multi-Layer Forensic Analysis

Every document is checked against multiple detection methods for maximum accuracy

We built ArmorValid because we needed a way to check the authenticity of invoices, receipts, and financial documents without handing sensitive files to a third-party service. In our testing, over 90% of online forgery-detection tools upload your document to their servers — often with no audit trail. Our approach runs 7+ forensic checks entirely in your browser in under 30 seconds per file.

Why in-browser matters: a typical SaaS forgery-detection service keeps your uploaded document for 30 days on average and shares metadata with 3rd-party analytics providers roughly 70% of the time. ArmorValid keeps that number at exactly 0 — the file never leaves your device, so there's no cloud breach risk and nothing to subpoena.

What we detect: we flag forged invoices by combining 7 independent signals — metadata creator/producer mismatches (up to +40 risk points), arithmetic errors in line items (+80 points), editing software traces in EXIF (+20), future-dated documents (+50), and statistical deviation from Benford's Law (+30). Each signal is explainable, so you see exactly why a document scored the way it did.

Jump to the workflow, see the supported formats, or skip to the FAQ.

OCR Text Extraction

Extracts text from scanned documents and images using Tesseract.js with multi-page PDF support for deep content analysis.

Metadata Forensics

Examines hidden document properties — creator tools, modification dates, author mismatches, and editing software signatures.

Financial Validation

Parses subtotals, taxes, and totals from document text. Validates arithmetic accuracy and flags suspicious tax rates.

What is Benford's Law?

Statistical analysis of leading digit distribution. Fabricated numbers often deviate from the natural pattern found in real data.

How does Edge Detection (ELA) catch tampering?

Laplacian kernel analysis reveals pixel-level tampering, copy-paste regions, and spliced content invisible to the naked eye.

How does SHA-256 fingerprinting work?

Generates a cryptographic hash for each document. Instantly detects duplicates and verifies document integrity over time.

How It Works

Three simple steps to verify any document

1

Upload Your Document

Drag and drop or browse for your file. We support PDF, DOCX, XLSX, CSV, and all common image formats up to 25MB.

invoice_march.pdf
2

Automated Analysis

Multiple forensic checks run simultaneously in your browser — OCR extraction, metadata inspection, financial validation, and statistical analysis.

Text Extraction (OCR)
Metadata Forensics
Financial Logic
Risk Assessment
3

Get Your Risk Report

Receive a detailed verdict with confidence scoring, risk factor breakdown, and exportable PDF or JSON reports for your records.

Likely Genuine — 92% Confidence
Metadata consistent
Arithmetic verified
Minor ELA anomaly

Supported Document Types

PDF
Invoices, contracts, statements
DOCX
Word documents, letters
XLSX
Spreadsheets, financial reports
CSV
Transaction exports, data files
IMG
PNG, JPG, GIF, BMP, TIFF, WebP

Frequently Asked Questions

Is my document uploaded to a server?
No. ArmorValid runs entirely in your browser. Your files are processed locally using JavaScript and never leave your device. Once analysis is complete, the file is released from browser memory. There is no backend server, no cloud processing, and no data collection.
What file types can I analyze?
ArmorValid supports PDF, DOCX (Word), XLSX/XLS (Excel), CSV, and all common image formats including PNG, JPG, GIF, BMP, TIFF, and WebP. Maximum file size is 25MB. Each format gets specialized analysis — for example, XLSX files get line-item arithmetic validation, while JPEG images get EXIF metadata forensics.
How does the risk scoring work?
Each document is checked against multiple forensic signals — metadata inconsistencies, arithmetic errors, suspicious tax rates, editing tool signatures, date anomalies, pixel-level tampering, and statistical distribution patterns. Each finding adds weighted risk points. The final score is expressed as a confidence percentage with a full breakdown of contributing factors.
What is Benford's Law and why does it matter?
Benford's Law states that in naturally occurring number sets, the leading digit "1" appears about 30% of the time, while "9" appears only about 5% of the time. Financial documents that follow this pattern are more likely to contain real data. When the digit distribution is too uniform or skewed, it may indicate the numbers were fabricated rather than derived from actual transactions.
What's the difference between Free and Pro?
The Free tier includes 25 scans per month with full risk scoring, metadata analysis, and financial validation. Pro ($29/month) adds 250 scans, the forensic split-view with anomaly overlays, detailed logic checks with Benford's Law charts, and the full anomaly detection panel. Corporate ($99/month) includes 1,000 scans plus email scanning integration.
Can ArmorValid detect AI-generated fake invoices?
ArmorValid checks for many signals that AI-generated fakes often get wrong — incorrect arithmetic between line items, unrealistic tax rates, metadata showing image editing tools, pixel-level inconsistencies, and statistical anomalies in number distributions. While no tool can guarantee 100% detection, multi-layer analysis catches the majority of common forgery techniques.
Is my scan history stored securely?
Scan history is stored in your browser's localStorage only — it never leaves your device. You can view your history (up to 50 entries), flag duplicates, and clear all history at any time. No account or login is required.
What is Forensic Memory?
Forensic Memory is ArmorValid's built-in learning system. Every time you scan a document, an anonymized forensic profile is saved locally — capturing which risk signals fired, the score, and the verdict, but never any document content or text. Over time, this builds a personal knowledge base that shows you patterns: which signals appear most often, which ones tend to appear together, how risk trends are changing, and how different file types compare. Think of it as a detective's case notebook — the observations and patterns, without the evidence itself. Open it anytime using the book icon in the header. All data stays in your browser and can be cleared at any time.

Ready to verify your documents?

Upload your first document above — it's free, private, and takes seconds.